Ningde era, a name that new energy car companies at home and abroad can’t get around. The 2022 annual performance forecast was recently announced, and the single-quarter profit exceeded 10 billion yuan for the first time, which made it a hot topic in the market.
The achievements in financial data are inseparable from the lithium battery layout of CATL over the years. Public information shows that the company currently has 13 lithium battery production bases. However, as the scale of production continued to expand, some problems were gradually exposed.
Ni Jun, Chief Manufacturing Officer of Ningde Times and President of Engineering Manufacturing System, once disclosed the challenges faced by lithium battery manufacturing: complex production process, huge material scale span, high safety requirements and quality consistency requirements, continuous performance improvement, high cost pressure, etc. wait.
Take the core production process as an example. Generally speaking, the production process of lithium batteries is divided into front-end process (electrode manufacturing), middle-end process (cell synthesis), and back-end process (chemical formation and packaging).
In the front-stage process and middle-stage winding, due to the support of high-speed online production, after each process is produced, the human eye cannot detect it, so it is necessary to use automated equipment for intelligent quality inspection; while in the middle-stage, back-stage process and packaging process, the early stage of the industry Using manual naked eye inspection is now facing problems such as long recruitment and training process and instability of manual inspection.
In order to meet the high requirements for safety performance of lithium-ion batteries, Ningde Times introduced industrial AI quality inspection technology into the safety judgment link in several processes of power batteries, and improved production efficiency by using AI analysis to comprehensively judge product quality inspection results.
This is not an isolated case. With the advent of the digital age, traditional manual inspection methods have become increasingly difficult to adapt to the development demands of enterprise digitalization. More accurate and stable industrial AI quality inspection solutions are being integrated into the production process of manufacturing products at an accelerated rate.
According to the “Market Analysis of China’s AI-Enabled Industrial Quality Inspection Solutions, 2022” report released by the international data company IDC, industrial AI quality inspection has moved from pilot application in the past few years to large-scale replication and promotion. Among them, the four industries of communication and electronics manufacturing, automobiles and parts, consumer goods and raw materials are the main application industries of industrial AI quality inspection at present, accounting for 91.5% of the market space in total.
And as this scene attracts various manufacturers to enter, the small market has become crowded.
01 The two-way rush of the industry
Industrial manufacturing is a typical representative of a country’s comprehensive strength. my country is the largest manufacturing country in the world, with 41 industrial categories, 207 industrial medium categories, and 666 industrial subcategories. It is the only country in the world that has all the industrial categories listed in the United Nations industrial classification.
But at the same time, there is still a long way to go to transform the domestic manufacturing industry to high-end, intelligent and green. Especially considering that in the production process, there are many types of parts and components, large quantities, and fast update iterations. There are high requirements for the accuracy, flexibility, and cost control of appearance inspections. Industrial quality inspection has become an indispensable link.
Traditional industrial quality inspection is mainly based on manual quality inspection. That is to say, it mainly relies on manpower, and quality inspectors need to have keen eyesight and rich experience to detect the type of defect. This model has very obvious disadvantages. At present, manufacturing enterprises generally face the problem of difficulty in recruiting and retaining employees. However, quality inspection is a hard job that requires a technical foundation, and there is a large gap in quality inspection staff. In addition, the detection speed is also limited, and there are hidden dangers of human judgment errors due to the influence of naked eye damage and fatigue.
“Manual quality inspection is the place where the most people are employed, and it faces obvious problems. On the one hand, the working hours are long and boring. Manually rotate the product 360 degrees to find defects. Each inspection takes up to 1 minute, and the staff is prone to fatigue. On the other hand On the one hand, it is impossible to accurately adapt personnel when orders change, and labor costs continue to increase, and the number of people employed during the peak period of quality inspection exceeds 1,500.” Deng Shengzhi, director of automation at Shanghai Fuchi Hi-Tech, once publicly stated.
On this basis, quality inspection methods based on traditional machine vision algorithms are widely used. Through machine vision, it is usually possible to process features that are easy to extract and quantify, such as color, area, roundness, rectangularity, angle, length, etc. Its principle is to perform logical judgment based on whether the target feature exists in the image or not, and the distance between multiple target features to complete the visual task.
This method, while replacing human labor, can satisfy inspection or measurement needs with relatively well-defined characteristics, and it can perform reliably when handling consistent and well-manufactured parts. But as the defect pool grows, the algorithm becomes more and more challenging.
Not only that, in the process of production line change and process upgrade, iterative learning is impossible; new defects and new features require new designs, and manual adjustment of various parameters is often required. The efficiency of algorithm development and debugging is low and the cycle is long, which has obvious bottlenecks.
In this context, the visual defect detection based on AI algorithm not only realizes the identification and detection of random defects, but also expands the application range of traditional machine vision, which has attracted widespread attention in the industry. Currently, the AI-powered industrial quality inspection software and solution products in the market mainly provide software/platforms, testing equipment, and customized testing systems for specific business scenarios. To some extent, this is also a two-way choice between AI quality inspection service providers and industry manufacturers.
From the perspective of industry manufacturers, Cheng Yin, senior analyst of artificial intelligence at IDC China, believes that during the implementation of AI-powered industrial quality inspection scenarios, because the factory’s business and technical leaders can identify defects based on the accuracy and error The inspection rate, reducing the amount of internal labor costs of the enterprise, and the timeliness of training and identifying defects are used to measure the business results of this scenario. Therefore, the AI industrial quality inspection scenario has been tried a lot in industrial enterprises due to its clear ROI, and has become a relatively mature field of industrial intelligence. application.
For AI quality inspection service providers, the importance of a scenario that can be implemented and scaled does not need to be repeated.
The China National Finance Securities Research Report once pointed out that the coldness of AI companies in the capital market is mainly related to the high expectations of the market in the early stage and the bottleneck in the development of the industry. The deep algorithm has not seen a breakthrough for a long time, the actual application scenarios are scattered, the degree of product standardization is low, the labor cost is high, and there are also challenges in morality and ethics, which make the business model and liquidity of AI algorithm companies challenged.
Due to the occasional meeting, the two sides have increasingly reached a consensus on promoting the implementation of AI quality inspection scenarios.
02 Cloud vendors dominate
At present, various manufacturers have entered the AI quality inspection market with their own basic advantages. Especially in the field of software and solutions, cloud vendors, AI startups, traditional machine vision companies, industrial Internet platform companies, etc., are all actively deploying in the field of AI visual quality inspection.
IDC’s report shows that in terms of AI quality inspection market share, the top five major manufacturers will have a market share of 44.3% in 2021, mainly concentrated in Baidu Smart Cloud, Chuangxin Qizhi, Tencent Cloud, Huawei, Aqiu Technology and other manufacturers.
It can be seen that cloud vendors are the main force in the AI industrial quality inspection software and solution market. To some extent, cloud vendors already have certain advantages in industrial quality inspection platforms, algorithm research and development, and data accumulation, and have accumulated the know-how of AI industrial vision implementation.
The reason why they are able to take the lead is related to their entry nodes and strategic deployment. Take Baidu as an example. As early as the 2017 Baidu Yunzhi Summit, 10,000 steel photos of Beijing Shougang Automation Information Technology Co., Ltd. were used to detect steel defects through the ABC all-in-one machine released by Baidu Smart Cloud. The detection results are very close, which is Baidu’s earliest attempt in the field of AI quality inspection.
Like Baidu, cloud vendors who are particularly concerned about AI quality inspection also include Tencent Cloud. At this stage, in the industrial AI quality inspection scene, Tencent Cloud has reached cooperation with benchmark customers such as Ningde Times and Shanghai Fuchi Hi-Tech, and can achieve large-scale replication.
At the end of 2022, the National Industrial Information Security Development Research Center of the Ministry of Industry and Information Technology will focus on key directions such as intelligent research and development, edge computing, intelligent production, intelligent operation, digital intelligent supply chain, and digital new infrastructure, soliciting excellent cases and projects with obvious value and benefits and demonstrable promotion from the society. Demonstration project. As one of Tencent Cloud’s smart products, its industrial AI quality inspection case successfully stood out and became the only Internet technology company case selected in the project list.
For Baidu and Tencent, the key to the AI quality inspection solution lies in the empowerment of the group’s cloud business.
In recent years, the growth driver of the cloud computing industry is switching from Internet customers to traditional enterprises. After several years of rapid development, the penetration rate of domestic Internet users has exceeded 70%, which means that without the stimulation of technological innovation, user traffic will gradually stabilize. On the contrary, traditional enterprise customers have gradually become the main force of migrating to the cloud. From the experience of the North American market, we can see that with the continuous advancement of digital transformation, traditional enterprises also have a relatively broad space for migrating to the cloud.
Therefore, it can be seen that Internet cloud vendors continue to expand in fields such as government affairs, industry, and energy. According to the “Economic Observation Network” report, in the non-Internet industry Tencent Cloud business is currently focusing on looking for opportunities in the energy industry and the manufacturing industry. The energy industry is relatively standardized, while the manufacturing industry is more complicated, but the advantage is that there are many business scenarios, and there is hope in individual production links. Achieve standardization and find larger application scenarios, such as AI quality inspection.
This is also more in line with the measurement and optimization of digitalization emphasized by Tang Daosheng, vice president of Tencent, that is, the core concept of Tencent’s industrial Internet activation industry-digitalization can be measured, and only with measurable data can optimization be carried out.
Of course, even with know-how, cloud vendors still need to spend a lot of effort in promoting their products. According to a former employee of Tencent Cloud, “Tencent focuses on promoting products such as manual quality inspection in the manufacturing and energy industries. No good breakthrough point was found.”
03 write at the end
According to IDC data, in the medium and long term, with the gradual differentiation of the AI quality inspection market and the improvement of application maturity, potential competition will lead to changes in the manufacturer structure, some manufacturers will expand their advantages, and new manufacturers will gradually join.
For example, Shangtang Technology, which is eager to find new application scenarios outside of security. On November 25, 2022, at the 2nd SenseTime Digital Energy Intelligent Manufacturing Ecological Partner Conference, SenseTime released a series of “AI intelligent quality inspection” software and hardware overall solutions.
It is conceivable that with the influx of a large number of manufacturers, the overall competition in industrial AI quality inspection has become increasingly fierce. It is reported that in the fields of electronics manufacturing and new energy vehicle power batteries, AI quality inspection manufacturers have launched a fierce price war.
On the other hand, the penetration rate of industrial AI quality inspection in the industry is still at a low level, and there is still significant room for development in the long run. IDC predicts that the overall market for China’s industrial AI quality inspection will reach US$958 million (approximately RMB 6.2 billion) in 2025, with a CAGR of 28.5% from 2021 to 2025.
At the same time, if it is considered that there are more than 2 million production line quality and efficiency related personnel on the production line of Chinese factories at this stage, the annual labor cost will consume 140 billion yuan; that is to say, the industrial AI quality inspection market has a market space of up to 100 billion yuan Whether it can better cope with the challenges of fragmented scenarios has become a new opportunity that manufacturers cannot ignore.
References:
China Industrial Internet Research Institute “Industrial AI Quality Inspection Standardization Research Report (2022)”
IDC “Market Analysis of China’s AI-Enabled Industrial Quality Inspection Solutions, 2022”